{"id":6203,"date":"2023-10-07T17:15:19","date_gmt":"2023-10-07T17:15:19","guid":{"rendered":"https:\/\/dailyai.com\/?p=6203"},"modified":"2023-10-07T22:49:19","modified_gmt":"2023-10-07T22:49:19","slug":"can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior","status":"publish","type":"post","link":"https:\/\/dailyai.com\/da\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","title":{"rendered":"Kan \"konstitutionel AI\" l\u00f8se problemet med problematisk AI-adf\u00e6rd?"},"content":{"rendered":"

I takt med at AI-modellerne bliver mere og mere integreret i vores hverdag, vokser bekymringen over begr\u00e6nsningerne og p\u00e5lideligheden af deres s\u00e5kaldte \"v\u00e6rn\".<\/b><\/p>\n

Allestedsn\u00e6rv\u00e6rende AI-modeller som GPT-3.5\/4\/4V m.fl. har indbyggede v\u00e6rn og sikkerhedsforanstaltninger, der forhindrer dem i at producere ulovlige, uetiske eller p\u00e5 anden m\u00e5de u\u00f8nskede output. <\/span><\/p>\n

Disse sikkerhedsfunktioner er dog langt fra uigennemtr\u00e6ngelige, og modeller viser deres potentiale til at l\u00f8srive sig fra deres gel\u00e6nder - eller k\u00f8re af sporet, s\u00e5 at sige. <\/span><\/p>\n

En del af problemet er, at gel\u00e6nderne ikke holder trit med modellernes kompleksitet og mangfoldighed.\u00a0<\/span><\/p>\n

I de seneste uger har OpenAI, som st\u00f8ttes af Microsoft, afsl\u00f8ret store forbedringer i ChatGPT, som g\u00f8r det muligt at interagere udelukkende ved hj\u00e6lp af stemmen og svare p\u00e5 foresp\u00f8rgsler via billeder og tekst. Denne multimodale billedkompatible version af GPT-4 er blevet d\u00f8bt \"GPT-4V\".<\/span><\/p>\n

Samtidig annoncerede Meta udrulningen af en AI-assistent, flere Ber\u00f8mte chatbot-personligheder<\/a> til WhatsApp- og Instagram-brugere og en r\u00e6kke andre lavm\u00e6lte AI-funktioner som AI Stickers.\u00a0<\/span><\/p>\n

Folk manipulerede straks Meta's <\/span>AI-klisterm\u00e6rker til at generere<\/span><\/a> komiske og chokerende tegneserielignende billeder, s\u00e5som Karl Marx n\u00f8gen eller Mario med et stormgev\u00e6r.\u00a0<\/span><\/p>\n

I takt med at kapl\u00f8bet om at kommercialisere AI intensiveres, viser det sig, at de sikkerhedsforanstaltninger, der er designet til at kontrollere AI-adf\u00e6rd - og forhindre dem i at generere skadeligt indhold, misinformation eller hj\u00e6lpe med ulovlige aktiviteter - bliver svagere.\u00a0<\/span><\/p>\n

Er konstitutionel AI svaret?<\/span><\/h2>\n

For at bek\u00e6mpe dette str\u00e6ber AI-udviklervirksomheder efter at skabe \"AI-forfatninger\", et s\u00e6t grundl\u00e6ggende principper og v\u00e6rdier, som AI-modeller skal overholde. Den nystartede virksomhed Antropisk<\/a> var blandt de f\u00f8rste til at g\u00e5 ind for 'konstitutionel AI' i en <\/span>2022 papir<\/span><\/a>.<\/span><\/p>\n

Google DeepMind etablerede ogs\u00e5 forfatningsm\u00e6ssige regler for sin chatbot Spurv i 2022<\/a> at f\u00f8re \"hj\u00e6lpsomme, korrekte og harml\u00f8se\" samtaler.\u00a0<\/span><\/p>\n

Anthropics AI-forfatninger udleder principper fra forskellige kilder, herunder FN's menneskerettighedserkl\u00e6ring og Apples servicevilk\u00e5r. Modellen er udstyret med grundl\u00e6ggende moralske principper, der driver adf\u00e6rden nedefra og op i stedet for at p\u00e5l\u00e6gge gel\u00e6ndere oppefra og ned.\u00a0<\/span><\/p>\n

I stedet for m\u00f8jsommeligt at tr\u00e6ne AI med utallige menneskelige eksempler p\u00e5 rigtigt og forkert, indlejrer denne tilgang et s\u00e6t regler eller principper - en \"forfatning\" - som AI'en f\u00f8lger.<\/p>\n

I f\u00f8rste omgang introduceres AI'en til en situation, derefter opfordres den til at kritisere sit svar, og til sidst finjusteres dens adf\u00e6rd baseret p\u00e5 den reviderede l\u00f8sning.<\/p>\n

Dern\u00e6st dykker systemet ned i forst\u00e6rkningsindl\u00e6ringsfasen. Her m\u00e5ler det kvaliteten af sine egne svar og skelner mellem de bedste. Over tid forbedrer denne selvevaluering dets adf\u00e6rd.<\/p>\n

Det nye er, at AI'en bruger sit feedback-loop til at bestemme bel\u00f8nningen i en metode, der kaldes \"RL from AI Feedback\" (RLAIF). N\u00e5r AI'en konfronteres med potentielt skadelige eller vildledende foresp\u00f8rgsler, g\u00e5r den ikke bare udenom eller afviser. I stedet g\u00e5r den direkte til sagen og forklarer, hvorfor en s\u00e5dan anmodning kan v\u00e6re problematisk.<\/p>\n

Det er et skridt fremad mod at skabe maskiner, der ikke kun regner, men ogs\u00e5 't\u00e6nker' p\u00e5 en struktureret m\u00e5de.<\/p>\n

Dario Amodei, CEO og medstifter af Anthropic, understregede udfordringen med at forst\u00e5, hvordan AI-modeller fungerer. Han foresl\u00e5r, at en forfatning vil g\u00f8re reglerne gennemsigtige og eksplicitte og sikre, at alle brugere ved, hvad de kan forvente.\u00a0<\/span><\/p>\n

Det er vigtigt, at det ogs\u00e5 giver mulighed for at holde modellen ansvarlig, hvis den ikke overholder de skitserede principper.<\/span><\/p>\n

P\u00e5 trods af disse bestr\u00e6belser er AI-forfatningerne ikke uden egne fejl, og modeller fra udviklere som Anthropic har vist sig at v\u00e6re s\u00e5rbare over for <\/span>jailbreaks<\/span><\/a> som s\u00e5 mange andre.\u00a0<\/span><\/p>\n

Der er ingen universelt accepterede veje til at tr\u00e6ne sikre og etiske AI-modeller<\/span><\/h2>\n

Historisk set er AI-modeller blevet forfinet ved hj\u00e6lp af en metode, der kaldes reinforcement learning by human feedback (RLHF), hvor AI-svar kategoriseres som \"gode\" eller \"d\u00e5rlige\" af store hold af menneskelige evaluatorer.\u00a0<\/span><\/p>\n

Selv om metoden er effektiv til en vis grad, er den blevet kritiseret for sin manglende n\u00f8jagtighed og specificitet. For at sikre AI-etik og -sikkerhed unders\u00f8ger virksomheder nu alternative l\u00f8sninger.<\/span><\/p>\n

OpenAI har f.eks. valgt en \"red-teaming\"-tilgang, hvor de ans\u00e6tter eksperter p\u00e5 tv\u00e6rs af forskellige discipliner til at teste og identificere svagheder i deres modeller.<\/span><\/p>\n

OpenAI's system fungerer i iterationer: AI-modellen producerer output, menneskelige anmeldere vurderer og korrigerer disse output ud fra specifikke retningslinjer, og modellen l\u00e6rer af denne feedback. Tr\u00e6ningsdataene fra disse anmeldere er afg\u00f8rende for modellens etiske kalibrering.<\/p>\n

ChatGPT v\u00e6lger ofte et konservativt svar, n\u00e5r den st\u00e5r over for kontroversielle eller f\u00f8lsomme emner, og undg\u00e5r nogle gange et direkte svar. Dette st\u00e5r i kontrast til konstitutionel AI, hvor modellen b\u00f8r belyse sine forbehold, n\u00e5r den pr\u00e6senteres for potentielt skadelige foresp\u00f8rgsler, og aktivt demonstrere r\u00e6sonnementer baseret p\u00e5 dens grundl\u00e6ggende regler.<\/p>\n

Mens ChatGPT i h\u00f8j grad er afh\u00e6ngig af menneskelig feedback for sin etiske orientering, bruger konstitutionel AI en regelbaseret ramme med mekanismer til selvrevision og v\u00e6gt p\u00e5 gennemsigtig argumentation.<\/p>\n

I sidste ende er der sandsynligvis ikke nogen universel tilgang til at udvikle 'sikker' AI - og nogle, som Elon Musk, kritiserer forestillingen om renset 'v\u00e5gen' AI. Unders\u00f8gelser har vist<\/a> at selv konstitutionelle AI'er kan jailbreakes og manipuleres til at opf\u00f8re sig uforudsigeligt.\u00a0<\/span><\/p>\n

Rebecca Johnson, der forsker i AI-etik ved University of Sydney, p\u00e5pegede, at AI-ingeni\u00f8rer og dataloger ofte n\u00e6rmer sig problemer med det form\u00e5l at finde endelige l\u00f8sninger, som m\u00e5ske ikke altid tager h\u00f8jde for den menneskelige naturs kompleksitet.\u00a0<\/span><\/p>\n

\"Vi er n\u00f8dt til at begynde at behandle generativ AI som udvidelser af mennesker, de er bare et andet aspekt af menneskeheden,\" sagde hun.\u00a0<\/span><\/p>\n

Omfattende kontrol med AI som en slags simpelt teknisk system vil kun blive <\/span>sv\u00e6rere, efterh\u00e5nden som den udvikler sig<\/span><\/a>Og det samme kan siges om biologiske organismer som os selv. <\/span><\/p>\n

Divergens, fremprovokeret eller ej, er m\u00e5ske uundg\u00e5elig. <\/span><\/p>","protected":false},"excerpt":{"rendered":"

I takt med at AI-modeller bliver mere og mere integreret i vores dagligdag, vokser bekymringen over begr\u00e6nsningerne og p\u00e5lideligheden af deres s\u00e5kaldte \"v\u00e6rn\". Allestedsn\u00e6rv\u00e6rende AI-modeller som GPT-3.5\/4\/4V m.fl. har indbyggede v\u00e6rn og sikkerhedsforanstaltninger, der skal forhindre dem i at producere ulovlige, uetiske eller p\u00e5 anden m\u00e5de u\u00f8nskede output. Disse sikkerhedsfunktioner er dog langt fra uigennemtr\u00e6ngelige, og modellerne viser sig at have potentiale til at l\u00f8srive sig fra deres r\u00e6kv\u00e6rk - eller k\u00f8re af sporet, s\u00e5 at sige. En del af problemet er, at gel\u00e6nderne ikke holder trit med modellernes kompleksitet og mangfoldighed. I de seneste uger har OpenAI, der st\u00f8ttes af Microsoft, afsl\u00f8ret store forbedringer<\/p>","protected":false},"author":2,"featured_media":6204,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[84],"tags":[148,422,93],"class_list":["post-6203","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-industry","tag-anthropic","tag-constitutional-ai","tag-openai"],"yoast_head":"\nCan "constitutional AI" solve the issue of problematic AI behavior? | DailyAI<\/title>\n<meta name=\"description\" content=\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called "guardrails" are mounting.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/dailyai.com\/da\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/\" \/>\n<meta property=\"og:locale\" content=\"da_DK\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Can "constitutional AI" solve the issue of problematic AI behavior? | DailyAI\" \/>\n<meta property=\"og:description\" content=\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called "guardrails" are mounting.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/dailyai.com\/da\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/\" \/>\n<meta property=\"og:site_name\" content=\"DailyAI\" \/>\n<meta property=\"article:published_time\" content=\"2023-10-07T17:15:19+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-10-07T22:49:19+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1000\" \/>\n\t<meta property=\"og:image:height\" content=\"667\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Sam Jeans\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:site\" content=\"@DailyAIOfficial\" \/>\n<meta name=\"twitter:label1\" content=\"Skrevet af\" \/>\n\t<meta name=\"twitter:data1\" content=\"Sam Jeans\" \/>\n\t<meta name=\"twitter:label2\" content=\"Estimeret l\u00e6setid\" \/>\n\t<meta name=\"twitter:data2\" content=\"4 minutter\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"NewsArticle\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"},\"author\":{\"name\":\"Sam Jeans\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\"},\"headline\":\"Can “constitutional AI” solve the issue of problematic AI behavior?\",\"datePublished\":\"2023-10-07T17:15:19+00:00\",\"dateModified\":\"2023-10-07T22:49:19+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"},\"wordCount\":915,\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"keywords\":[\"Anthropic\",\"Constitutional AI\",\"OpenAI\"],\"articleSection\":[\"Industry\"],\"inLanguage\":\"da-DK\"},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\",\"name\":\"Can \\\"constitutional AI\\\" solve the issue of problematic AI behavior? | DailyAI\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"datePublished\":\"2023-10-07T17:15:19+00:00\",\"dateModified\":\"2023-10-07T22:49:19+00:00\",\"description\":\"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called \\\"guardrails\\\" are mounting.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#breadcrumb\"},\"inLanguage\":\"da-DK\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"da-DK\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#primaryimage\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/10\\\/shutterstock_2276608417-1.jpg\",\"width\":1000,\"height\":667,\"caption\":\"Anthropic AI\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/2023\\\/10\\\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/dailyai.com\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Can “constitutional AI” solve the issue of problematic AI behavior?\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#website\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"name\":\"DailyAI\",\"description\":\"Your Daily Dose of AI News\",\"publisher\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/dailyai.com\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"da-DK\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#organization\",\"name\":\"DailyAI\",\"url\":\"https:\\\/\\\/dailyai.com\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"da-DK\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"contentUrl\":\"https:\\\/\\\/dailyai.com\\\/wp-content\\\/uploads\\\/2023\\\/06\\\/Daily-Ai_TL_colour.png\",\"width\":4501,\"height\":934,\"caption\":\"DailyAI\"},\"image\":{\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/logo\\\/image\\\/\"},\"sameAs\":[\"https:\\\/\\\/x.com\\\/DailyAIOfficial\",\"https:\\\/\\\/www.linkedin.com\\\/company\\\/dailyaiofficial\\\/\",\"https:\\\/\\\/www.youtube.com\\\/@DailyAIOfficial\"]},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/dailyai.com\\\/#\\\/schema\\\/person\\\/711e81f945549438e8bbc579efdeb3c9\",\"name\":\"Sam Jeans\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"da-DK\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g\",\"caption\":\"Sam Jeans\"},\"description\":\"Sam is a science and technology writer who has worked in various AI startups. When he\u2019s not writing, he can be found reading medical journals or digging through boxes of vinyl records.\",\"sameAs\":[\"https:\\\/\\\/www.linkedin.com\\\/in\\\/sam-jeans-6746b9142\\\/\"],\"url\":\"https:\\\/\\\/dailyai.com\\\/da\\\/author\\\/samjeans\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Kan \"konstitutionel AI\" l\u00f8se problemet med problematisk AI-adf\u00e6rd? | DailyAI","description":"I takt med at AI-modellerne bliver mere og mere integreret i vores dagligdag, vokser bekymringen over de s\u00e5kaldte \"gel\u00e6nders\" begr\u00e6nsninger og p\u00e5lidelighed.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/dailyai.com\/da\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","og_locale":"da_DK","og_type":"article","og_title":"Can \"constitutional AI\" solve the issue of problematic AI behavior? | DailyAI","og_description":"As AI models continue to embed themselves in our daily lives, concerns over the limitations and reliability of the so-called \"guardrails\" are mounting.","og_url":"https:\/\/dailyai.com\/da\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","og_site_name":"DailyAI","article_published_time":"2023-10-07T17:15:19+00:00","article_modified_time":"2023-10-07T22:49:19+00:00","og_image":[{"width":1000,"height":667,"url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","type":"image\/jpeg"}],"author":"Sam Jeans","twitter_card":"summary_large_image","twitter_creator":"@DailyAIOfficial","twitter_site":"@DailyAIOfficial","twitter_misc":{"Skrevet af":"Sam Jeans","Estimeret l\u00e6setid":"4 minutter"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"NewsArticle","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#article","isPartOf":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"},"author":{"name":"Sam Jeans","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9"},"headline":"Can “constitutional AI” solve the issue of problematic AI behavior?","datePublished":"2023-10-07T17:15:19+00:00","dateModified":"2023-10-07T22:49:19+00:00","mainEntityOfPage":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"},"wordCount":915,"publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"image":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","keywords":["Anthropic","Constitutional AI","OpenAI"],"articleSection":["Industry"],"inLanguage":"da-DK"},{"@type":"WebPage","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","url":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/","name":"Kan \"konstitutionel AI\" l\u00f8se problemet med problematisk AI-adf\u00e6rd? | DailyAI","isPartOf":{"@id":"https:\/\/dailyai.com\/#website"},"primaryImageOfPage":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"image":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage"},"thumbnailUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","datePublished":"2023-10-07T17:15:19+00:00","dateModified":"2023-10-07T22:49:19+00:00","description":"I takt med at AI-modellerne bliver mere og mere integreret i vores dagligdag, vokser bekymringen over de s\u00e5kaldte \"gel\u00e6nders\" begr\u00e6nsninger og p\u00e5lidelighed.","breadcrumb":{"@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#breadcrumb"},"inLanguage":"da-DK","potentialAction":[{"@type":"ReadAction","target":["https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/"]}]},{"@type":"ImageObject","inLanguage":"da-DK","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#primaryimage","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/10\/shutterstock_2276608417-1.jpg","width":1000,"height":667,"caption":"Anthropic AI"},{"@type":"BreadcrumbList","@id":"https:\/\/dailyai.com\/2023\/10\/can-constitutional-ai-solve-the-issue-of-problematic-ai-behavior\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/dailyai.com\/"},{"@type":"ListItem","position":2,"name":"Can “constitutional AI” solve the issue of problematic AI behavior?"}]},{"@type":"WebSite","@id":"https:\/\/dailyai.com\/#website","url":"https:\/\/dailyai.com\/","name":"DailyAI","description":"Din daglige dosis af AI-nyheder","publisher":{"@id":"https:\/\/dailyai.com\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/dailyai.com\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"da-DK"},{"@type":"Organization","@id":"https:\/\/dailyai.com\/#organization","name":"DailyAI","url":"https:\/\/dailyai.com\/","logo":{"@type":"ImageObject","inLanguage":"da-DK","@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/","url":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","contentUrl":"https:\/\/dailyai.com\/wp-content\/uploads\/2023\/06\/Daily-Ai_TL_colour.png","width":4501,"height":934,"caption":"DailyAI"},"image":{"@id":"https:\/\/dailyai.com\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/x.com\/DailyAIOfficial","https:\/\/www.linkedin.com\/company\/dailyaiofficial\/","https:\/\/www.youtube.com\/@DailyAIOfficial"]},{"@type":"Person","@id":"https:\/\/dailyai.com\/#\/schema\/person\/711e81f945549438e8bbc579efdeb3c9","name":"Sam Jeans","image":{"@type":"ImageObject","inLanguage":"da-DK","@id":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/a24a4a8f8e2a1a275b7491dc9c9f032c401eabf23c3206da4628dc84b6dac5c8?s=96&d=robohash&r=g","caption":"Sam Jeans"},"description":"Sam er videnskabs- og teknologiforfatter og har arbejdet i forskellige AI-startups. N\u00e5r han ikke skriver, kan han finde p\u00e5 at l\u00e6se medicinske tidsskrifter eller grave i kasser med vinylplader.","sameAs":["https:\/\/www.linkedin.com\/in\/sam-jeans-6746b9142\/"],"url":"https:\/\/dailyai.com\/da\/author\/samjeans\/"}]}},"_links":{"self":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/6203","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/comments?post=6203"}],"version-history":[{"count":12,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/6203\/revisions"}],"predecessor-version":[{"id":6216,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/posts\/6203\/revisions\/6216"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/media\/6204"}],"wp:attachment":[{"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/media?parent=6203"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/categories?post=6203"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dailyai.com\/da\/wp-json\/wp\/v2\/tags?post=6203"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}